DEV Community: Ali Farhat

Google Search Console Platform Properties Bring Social Query Data to Content Teams

Ali Farhat — Sat, 22 Aug 2026 13:15:30 +0000

Google Search Console is expanding its measurement scope beyond websites with platform properties, a new property type for linked social and video accounts. The feature lets eligible creators and site owners see how content from Instagram, TikTok, X and YouTube is discovered across Google Search, Discover and Google News, including the queries that drive traffic to posts and videos.

The important shift is not simply another social metric. Google is bringing search-demand signals for social content into the same product used to assess organic search performance. That gives SEO professionals, social teams and creators a clearer way to connect what people search for with the posts, videos and formats that receive Google-driven discovery.

In Google's announcement of Search Console platform properties, the company says linked accounts can be analyzed through Search Console's Performance and Insights reports, with export options available. Google is rolling out the feature gradually over the coming weeks, and each account must be added and verified in Search Console.

What platform properties add to Search Console

A platform property is designed for content hosted on a supported social or video platform rather than on a verified website. Once an account is linked, Search Console can show when content begins to gain traction, which posts or videos are leading performance, and which queries are sending people to that content from Google's surfaces.

That makes the feature materially different from a website-only view of organic traffic. A brand may publish a topic on its own site and also distribute related short-form video, social posts or creator content. Platform properties can help teams assess the Google discovery side of those channels without treating them as entirely separate editorial systems.

The currently named supported platforms are:

Instagram
TikTok
X
YouTube

Linked platform	Content covered by the platform property	Google discovery data described by Google
Instagram	Social content from a verified account	Queries, leading content and traffic patterns across Google surfaces
TikTok	Social and video content from a verified account	Queries, leading content and traffic patterns across Google surfaces
X	Social content from a verified account	Queries, leading content and traffic patterns across Google surfaces
YouTube	Video content from a verified account	Queries, leading content and traffic patterns across Google surfaces

Google's accompanying documentation says users can apply filters and grouping to study traffic patterns. The practical value is the ability to move from a broad observation, such as a video receiving more Google exposure, to a more useful question: which search terms and content assets are associated with that discovery?

For editorial teams, that can inform creator briefs and cross-promotion decisions. A query pattern that consistently leads to a particular post or video may point to a topic worth covering in another format. It may also help teams evaluate captions, hashtags and content framing with a clearer view of search-led discovery, rather than relying only on in-platform performance reporting.

Implications for content operations and governance

Search Console platform properties create a shared measurement layer for teams that often work in separate tools and reporting structures. SEO teams can identify search demand. Social managers can see which social assets are appearing in Google-driven journeys. Content leads can use the combined evidence to decide whether a topic needs a new post, video, supporting article or a more focused creator brief.

The rollout also introduces operational requirements. Because each social or video account must be added and verified, organizations need to know who owns access to each account and who is responsible for maintaining the connection. That is a governance issue as much as a setup task, particularly for companies with regional accounts, multiple business units or agency-managed channels.

A disciplined early-adoption workflow could include:

Inventorying the Instagram, TikTok, X and YouTube accounts that the organization controls.
Assigning ownership for verification and access management in Search Console.
Reviewing query and top-content data alongside existing website and social reporting.
Using recurring search themes to shape briefs, cross-promotion and content-format decisions.

The feature is also relevant to SEO tooling and reporting design. Search Console exports allow teams to incorporate the data into their existing analysis process, but the initial focus should be on clear questions rather than volume alone. For example, teams can examine whether a query is associated with a post that is gaining visibility, whether related content should be developed elsewhere, and whether the organization has a gap in its coverage of a searched topic.

There are limits to keep in view. The rollout is gradual, so availability may differ during the launch period. Account verification is required, and the available data is subject to platform API and privacy controls. Organizations should therefore treat early reports as a new evidence source to validate against their established analytics, not as a complete replacement for native platform reporting or website measurement.

For businesses publishing across search and social, this update makes measurement design more important: teams need accountable account ownership, repeatable query analysis, and briefs that turn discovery data into useful content before reporting processes become fragmented. Scalevise can help connect those decisions to broader AI-search visibility, showing where brand answers and formats need attention. Start an AI Visibility scan with Scalevise to prioritize the next content opportunities.

Frequently Asked Questions

What are Google Search Console platform properties?

Google Search Console platform properties are a new property type for linking supported social and video accounts so users can analyze how their content performs across Google Search, Discover and Google News.

Which platforms can be linked to Search Console platform properties?

Google has named Instagram, TikTok, X and YouTube as supported platforms for platform properties.

What data can users see after linking a platform property?

Google says users can see the queries driving traffic to social content, the posts or videos performing best, when content starts taking off, and traffic patterns using filters and grouping.

Where does the platform-property data appear in Search Console?

Google says the data is available through the Performance and Insights reports, with export options available for further analysis.

Is the feature available to every account immediately?

No. Google says platform properties are rolling out gradually over the coming weeks, and each supported account must be added and verified.

Conclusion

Google Search Console platform properties give social and video content a more direct place in search performance analysis. By exposing the queries and Google surfaces associated with discovery, the feature can help teams connect SEO research, social publishing and creator workflows. Its value will depend on careful account verification, sensible reporting practices and an understanding of the data available as the gradual rollout progresses.

How EU-Funded AI and Organ-on-Chip Research Could Reshape Chemical Safety Testing

Ali Farhat — Sat, 22 Aug 2026 10:00:30 +0000

European researchers are building a more human-relevant approach to chemical safety testing by combining artificial intelligence, human cells and organ-on-chip models. The EU-backed work aims to improve predictions of how chemicals may affect people while reducing dependence on animal experiments. It does not promise an immediate end to animal testing, but it marks a coordinated attempt to make non-animal evidence more useful in toxicology and regulatory decision-making.

The work is organised through the ASPIS cluster, an approximately €60 million group of projects focused on next-generation chemical safety testing. As outlined in the European Commission's Horizon Magazine report on AI and animal testing, the cluster brings together complementary methods: AI-supported toxicity prediction, cross-species testing and the integration of laboratory findings with exposure data from the real world.

This matters because chemical safety assessment is not simply a laboratory challenge. For a non-animal method to influence formal decisions, its results must be dependable, interpretable and usable across organisations. The research therefore has implications beyond biotech. It highlights the data governance, validation and interoperability requirements that enterprises will face as AI becomes part of regulated scientific workflows.

From animal studies to new approach methodologies

The EU effort sits within the 3Rs framework: Replacement, Reduction and Refinement of animal use in science. Its practical focus is on new approach methodologies, often called NAMs. These methodologies can include human-cell systems, computational models, organ-on-chip technologies and other non-animal approaches that generate evidence about biological effects.

Organ-on-chip models are particularly relevant because they use human cells in systems designed to model aspects of organ function. Combined with AI, the resulting data may help researchers identify patterns associated with toxic effects and assess results across complex datasets. The objective is not merely to generate more data. It is to create evidence that better reflects human biology and can support chemical safety assessment.

ASPIS has been active since 2021 and includes three linked initiatives with distinct roles:

Initiative	Primary focus	Role in chemical safety research
ONTOX	AI-driven toxicity testing for kidney, liver and brain	Develops approaches for predicting toxicity in key human-relevant organ systems.
PrecisionTox	Cross-species testing	Compares responses across five model organisms and human cells for around 200 chemicals.
RISK-HUNT3R	Risk assessment integration	Combines non-animal test results with real-world exposure data to assess risk.

Together, these projects address different parts of the same problem. ONTOX focuses on toxicity signals in specific organs, PrecisionTox explores what can be learned by comparing biological responses across species and human cells, and RISK-HUNT3R links experimental evidence to the exposure conditions that shape real-world risk. That combination is important: a laboratory signal alone is not equivalent to an assessment of human risk.

Why AI is useful, and where it is not enough

AI can help analyse the varied data generated by cell-based tests, organ-on-chip models and other NAMs. In principle, it can identify relationships that would be difficult to detect manually and support more consistent toxicity predictions. Yet an AI output does not automatically become regulatory-grade evidence.

Several conditions will determine whether these methods can move into formal safety pipelines:

Standards and validation are needed to establish when a method produces reliable, reproducible results.
Data interoperability is essential if results from different laboratories, systems and projects are to be combined meaningfully.
Clear governance is required to document methods, inputs and decision processes in regulated settings.
Regulatory acceptance remains central, because promising scientific methods must still fit the evidence requirements used in chemical safety evaluations.

The Horizon Magazine report notes growing scientific and regulatory engagement, including interest from the OECD. That engagement is significant because common approaches and accepted methods can determine whether research advances remain confined to projects or become usable at scale.

The regulatory challenge is as important as the model

The research is directly relevant to the EU's REACH framework for chemical regulation. REACH creates a substantial need for evidence on chemical hazards and safety, while the 3Rs framework creates pressure to find credible alternatives to animal testing. NAMs could help reconcile those needs, but only if their results can be assessed with appropriate confidence.

The current position is more measured than a claim that animal testing is about to disappear. The researchers' aim is to replace or substantially reduce animal testing, while recognising that complete abolition is not yet feasible. That distinction matters for companies, policymakers and technology providers. The near-term opportunity is likely to be the targeted adoption of validated non-animal methods within broader assessment workflows, rather than a wholesale switch from one evidence system to another.

For biotech organisations, this places equal emphasis on biology and operational design. Teams need systems that preserve data quality, record provenance and allow results to be reviewed across research, safety and compliance functions. For enterprise AI teams, the lesson is similar: model performance alone is insufficient where outputs contribute to high-stakes decisions. Traceability, interoperable data structures and well-defined human oversight must be designed into the workflow.

As AI-supported NAMs mature, the organisations best positioned to benefit will be those that treat governance as a capability, not a late-stage compliance exercise. Building trusted pipelines early can make it easier to evaluate new models, connect experimental and exposure data, and respond when regulators establish clearer expectations.

For biotech and chemical-sector leaders, this shift raises a practical question: can your AI and data architecture support evidence that scientists, compliance teams and regulators can scrutinise? Scalevise helps organisations translate emerging AI requirements into governed implementation plans, from workflow design to oversight and integration. A focused AI consultancy conversation with Scalevise can identify where your current tooling creates risk or slows adoption, and where a more auditable approach can create value. Request a consultation to assess your AI governance readiness.

Frequently Asked Questions

What is the ASPIS cluster?

ASPIS is an approximately €60 million EU-backed cluster of projects developing next-generation approaches to chemical safety testing. Its linked initiatives include ONTOX, PrecisionTox and RISK-HUNT3R.

How do organ-on-chip models support chemical safety testing?

Organ-on-chip models use human cells in systems intended to model aspects of organ function. They can provide human-relevant biological data that researchers may use alongside AI and other non-animal methods.

Will AI and organ-on-chip models fully replace animal testing?

Not yet. The EU research aims to replace or substantially reduce animal testing, but the supplied research notes that complete abolition is not currently feasible.

What are new approach methodologies, or NAMs?

NAMs are non-animal methods used to generate evidence about chemical effects. They can include human-cell systems, organ-on-chip models, computational approaches and other techniques.

Why do standards and data interoperability matter for NAMs?

Formal chemical safety evaluation requires reliable and interpretable evidence. Standards, validation and interoperable data help make results comparable, reproducible and usable across laboratories, organisations and regulatory processes.

Conclusion

The ASPIS projects show that AI-enabled toxicology is becoming a coordinated scientific and regulatory effort, not a standalone technology experiment. Human-cell and organ-on-chip models may make chemical testing more relevant to people and less reliant on animals, but their wider impact depends on validation, shared standards and trusted data governance. The next milestone is not simply better models. It is turning their evidence into methods that can support real chemical safety decisions.

OpenAI Cuts GPT-5.6 Luna and Terra Costs, Reshaping API Budget Planning

Ali Farhat — Sat, 22 Aug 2026 08:45:30 +0000

OpenAI has reduced usage costs for two GPT-5.6 model variants, cutting GPT-5.6 Luna pricing by about 80% and GPT-5.6 Terra pricing by about 20%. The changes apply to API usage and to the way credits are consumed in ChatGPT Work and Codex. At the same time, OpenAI introduced a Fast mode for GPT-5.6 Sol that can process work up to 2.5 times faster at roughly twice the price, while Sol's standard pricing remains unchanged.

The announcement matters because it changes the practical economics of deploying GPT-5.6 for high-volume workloads. Developers building copilots, internal tools, agent-based applications, and other token-intensive systems have a clearer lower-cost path through Luna and Terra. Teams that value response speed over unit cost can instead consider Sol Fast mode. OpenAI describes the lineup and its price-performance changes in its official GPT-5.6 overview.

What changed across the GPT-5.6 family

The update is not a blanket price cut for every GPT-5.6 option. It is a targeted adjustment to Luna and Terra, combined with a new performance tier for Sol. That distinction is important for procurement, forecasting, and model-routing policies: organizations should not assume that workloads currently assigned to standard Sol will automatically cost less.

GPT-5.6 option	Pricing or performance change	Practical implication
Luna	Pricing reduced by about 80%	Lower-cost option for high-volume usage
Terra	Pricing reduced by about 20%	Improved cost position for applicable workloads
Sol standard	Standard pricing unchanged	Existing base cost structure remains in place
Sol Fast mode	Up to 2.5x faster at roughly twice the price	A speed-focused option with a higher cost

For API users, the central change is the reduction in real token costs for Luna and Terra. For ChatGPT Work and Codex customers, the same update affects credit consumption. These are related but operationally distinct buying models, so teams should assess their own usage patterns rather than treating the percentages as a universal reduction in total AI spending.

The revised structure also makes model selection more explicitly a cost, throughput, and latency decision. Luna and Terra may become more attractive for recurring, high-volume processes where aggregate usage is the main budget concern. Sol Fast mode gives teams an option when faster processing has enough business value to justify its roughly doubled price.

Implications for developers and AI budgets

For startups, a substantial Luna reduction can alter the threshold at which a feature becomes affordable to run at scale. A product team that limited model calls, shortened workflows, or constrained internal testing because of token costs may be able to revisit those choices. The financial benefit will depend on how much of the application can appropriately use Luna rather than another model variant.

For enterprise teams, the announcement is a reason to refresh AI workload assumptions. Model costs often sit inside broader expenses such as retrieval systems, data processing, observability, human review, and application infrastructure. Lower token prices can improve a program's economics, but they do not remove the need to measure end-to-end cost and quality.

A practical review should focus on three areas:

Model routing: Identify which high-volume tasks can use Luna or Terra without compromising required output quality.
Credit and API forecasts: Recalculate usage projections separately for API deployments, ChatGPT Work, and Codex where relevant.
Governance controls: Preserve approval rules, spend monitoring, evaluation practices, and access controls as increased usage becomes more feasible.

The new Sol Fast mode adds another layer to that exercise. Its value is not lower standard Sol pricing, but a different price-performance point. A workflow with time-sensitive processing requirements may benefit from the higher-speed mode, while a batch-oriented system may prioritize Luna or Terra's lower cost. Organizations should validate the right choice against their own latency, quality, and volume requirements.

This is also a useful reminder that public model pricing is only one part of a deployment decision. Regional rollout, plan-specific eligibility, quotas, and the mechanics of credit use can affect the final cost and availability for a particular customer. Teams should confirm current terms in OpenAI's official pricing materials and developer documentation before changing production commitments.

For businesses expanding AI-enabled products, lower-cost model options can create room to test more use cases, but scaling without clear measurement can quickly obscure whether savings are real. Scalevise helps teams connect AI visibility, model choices, and commercial outcomes through an AI Visibility and GEO assessment. It can clarify where AI-driven experiences influence discovery and where better measurement should guide investment. Start an AI Visibility scan to prioritize the opportunities worth funding.

Frequently Asked Questions

What did OpenAI change in GPT-5.6 pricing?

OpenAI reduced GPT-5.6 Luna pricing by about 80% and GPT-5.6 Terra pricing by about 20%. The changes apply to API usage and credit consumption for ChatGPT Work and Codex.

Did OpenAI lower the standard price of GPT-5.6 Sol?

No. Standard GPT-5.6 Sol pricing remained unchanged. OpenAI added a Fast mode for Sol instead.

How does GPT-5.6 Sol Fast mode work?

Sol Fast mode offers up to 2.5 times faster processing at roughly twice the price. It is a speed-focused tradeoff, not a reduction in standard Sol pricing.

Why do the Luna and Terra reductions matter for developers?

Lower token costs can improve the economics of high-volume applications, including tooling, copilots, and agent-based workflows, where Luna or Terra meet the required workload needs.

Should enterprises immediately change their GPT-5.6 budgets?

Enterprises should review usage by model and purchasing channel first. Regional availability, plan eligibility, quotas, and credit consumption mechanics can affect actual costs.

Conclusion

OpenAI's GPT-5.6 update makes Luna and Terra more cost-competitive for applicable high-volume workloads, while Sol Fast mode creates a premium option for teams that prioritize processing speed. The practical opportunity is not a universal reduction across the family. It is a reason to reassess model routing, usage forecasts, and governance so AI spending aligns with each workload's cost and performance requirements.

Gemini’s Local Search Volatility Shows Why AI Visibility Needs Multi-Engine Measurement

Ali Farhat — Sat, 22 Aug 2026 01:15:30 +0000

Generative AI is becoming a new route into local business discovery, but its answers may be far less repeatable than established local search results. A Steady Demand field study of Gemini and ChatGPT found that Gemini returned the same top local-business recommendation only about 7% of the time when an identical prompt was repeated. By comparison, Google’s traditional Local Pack returned the same top listing roughly 90% of the time.

The result changes what an AI visibility metric can credibly represent. A single prompt response is not a durable ranking position. Instead, it is one observation from a variable system whose cited sources, leading recommendation, and source ecosystem can change across runs and differ substantially between AI engines.

What the local AI citation study found

Steady Demand’s August 6, 2026 study, What Gemini and ChatGPT Cite in Local Search, analyzed 1,487 identical local-intent prompts across 50 U.S. metropolitan areas and 10 service categories. The research produced 14,472 citations from Google Gemini and OpenAI ChatGPT.

The study describes this instability as Grounding Drift. Repeating the same query word for word produced overlapping cited sources only about 40% of the time. Gemini’s top-brand result was especially variable, even when the query itself did not change. Steady Demand’s methodology and interactive dashboards attribute the observed pattern to the AI generation process rather than changes in the underlying Google Business Profile or local data.

Measurement	Google Gemini	Google Local Pack
Consistency of the top local result on repeated identical queries	About 7%	Roughly 90%
Nature of the result	Generated local-business recommendation	Traditional local listing result

Grounding Drift makes one-off reporting unreliable

Traditional local-rank tracking is built around repeated observation of a comparatively stable result set. That approach does not translate directly to an AI answer surface where the leading business can change from one identical request to the next. Reporting that a brand ranked first in one Gemini response may be accurate for that response, yet misleading if presented as a stable position.

The practical unit of analysis should therefore be a distribution of outcomes, not an isolated answer. Teams need to know how often a business is cited, how often it leads recommendations, which domains appear alongside it, and how those measures vary by engine, market, prompt, and service category.

AI engines draw from different source ecosystems

The volatility is only part of the measurement challenge. The study found limited overlap between Gemini and ChatGPT: about 8% of cited domains overlapped, while only about 4.2% of top-1 brands overlapped. A business that appears prominently in one assistant should not assume it will receive comparable exposure in another.

Source patterns also varied by vertical. Gemini cited a business’s own website in 60% of citations, while Reddit and directories played outsized roles in different categories. Those findings argue against a universal optimization checklist. Local AI discovery depends on the sources each engine selects and how those source ecosystems behave for a particular service category.

For enterprise teams, the study supports several operational conclusions:

Test repeatedly because one response cannot establish a reliable visibility position.
Monitor multiple engines because Gemini and ChatGPT can surface different businesses and domains.
Segment reporting by metro, service category, prompt type, and engine rather than using a single blended score.
Separate observation from action by documenting sample sizes, prompt design, timing, and decision rules before changing content or local-search strategy.

Local AI discovery is becoming a measurement and governance problem, not simply an SEO reporting exercise. Scalevise helps teams establish repeatable, multi-engine visibility baselines, interpret volatile recommendation patterns, and connect findings to responsible content and operating decisions. Its AI Visibility / GEO Checker gives stakeholders a practical starting point for examining how brands appear across AI answers rather than treating one response as definitive. Start an AI Visibility scan.

How enterprises can govern AI visibility measurement

A useful governance model begins by defining what the organization is measuring. Citation presence, top recommendation frequency, source-domain representation, and consistency across repeat tests are distinct metrics. Combining them without clear definitions can obscure whether an apparent change reflects an actual pattern, a change in prompt construction, or normal generative variation.

Build a repeatable sampling framework

A repeatable framework should use standardized local-intent prompts and run them multiple times across the engines relevant to the business. Results should retain the prompt wording, engine, location, service category, run timing, cited domains, and recommendation position. This creates an auditable record and allows teams to compare distributions rather than anecdotes.

The aim is not to assume that AI answers will reproduce traditional search rankings. It is to understand the conditions under which a brand is represented, the sources associated with that representation, and the degree of uncertainty decision-makers should attach to the result.

Tie actions to evidence, not isolated outputs

Governance also requires thresholds for action. If a brand is absent from one AI response, that alone is weak evidence for a major content, directory, or local-listing intervention. Repeated patterns across relevant queries and engines provide a more defensible basis for prioritization.

This distinction matters for larger organizations managing multiple markets and categories. Without a common testing protocol, separate teams may draw conflicting conclusions from different AI responses. A documented measurement approach helps ensure that AI visibility reporting informs decisions without overstating what any individual generated answer means.

Frequently Asked Questions

What is Grounding Drift in local AI search?

Grounding Drift is Steady Demand’s term for the variation in AI citations and recommendations when the same local-intent query is repeated. In the study, identical prompts produced overlapping cited sources only about 40% of the time.

How consistent were Gemini’s top local recommendations?

Gemini returned the same top local-business recommendation only about 7% of the time across repeated identical queries in Steady Demand’s study.

How did Gemini compare with Google’s Local Pack?

Google’s traditional Local Pack returned the same top listing roughly 90% of the time on identical repeated queries, substantially more consistently than Gemini’s generated top recommendation.

Why should businesses measure more than one AI engine?

Gemini and ChatGPT showed limited citation overlap in the study, with about 8% of domains and about 4.2% of top-1 brands overlapping. Multi-engine measurement helps reveal engine-specific visibility patterns.

Conclusion

Steady Demand’s findings show that AI-driven local discovery cannot be assessed with the assumptions used for conventional rank tracking. Gemini’s volatile top recommendations, limited overlap with ChatGPT, and vertical-specific source patterns make repeated, multi-engine measurement essential. Businesses that treat AI visibility as a governed distribution of outcomes will have a firmer basis for interpreting results and deciding where to focus.

Waymo’s Gemini in Ojai Remains a Limited Beta as Rider Access Expands

Ali Farhat — Fri, 21 Aug 2026 21:45:30 +0000

Waymo’s Gemini integration in Ojai is being introduced through a limited beta rollout, not as a feature available to every rider. The company’s official updates describe early access for a restricted group of riders, followed by gradual expansion, while Gemini in the Ojai cabin continues to receive interface and product enhancements.

Waymo first outlined the rollout in May 2026, saying it would welcome initial riders in select cities, including San Francisco, Phoenix and Los Angeles. Those riders would receive free rides as Waymo collected feedback, with access expected to expand over time. Waymo’s May 2026 Ojai rollout announcement is the primary account of that initial plan.

The company’s July 2026 update characterized Gemini in Waymo as a beta feature in the Ojai cabin. It also referred to a redesigned user interface and ongoing enhancements. Taken together, the two announcements establish a phased deployment: early rider feedback first, followed by continued iteration and broader access over time.

What Waymo has announced about Gemini in Ojai

The official information supports a narrower view of availability than a universal launch. In May, Waymo described a limited group of early riders in three named cities. By July, the company was still describing Gemini in the Ojai cabin as a beta.

Rollout stage	Waymo’s description	What it indicates
May 2026	First riders in San Francisco, Phoenix and Los Angeles, with free rides for a limited group while Waymo collected feedback.	Initial access was restricted and feedback-led.
July 2026	Gemini in Waymo was described as a beta feature in the Ojai cabin, with a redesigned UI and ongoing enhancements.	The in-cabin experience remained in active development.

Waymo did not provide a universal-access date in the supplied material. Its stated approach is gradual expansion, which means availability should be assessed by the company’s current rider communications rather than by assuming that a beta feature has reached all users.

Why the beta designation matters

A beta label is more than a marketing qualifier. It signals that the experience is still being refined as the company gathers feedback and evolves the product. For autonomous-mobility operators, partners and technology teams, that distinction matters when assessing how quickly an in-cabin AI capability can move from a controlled rollout to a broadly available service.

The available official descriptions also set clear limits on what can be concluded. They identify Gemini as an Ojai cabin beta, note a redesigned UI and ongoing enhancements, and describe the initial rider-access model. They do not specify the complete set of Gemini interactions, a data-handling policy for the feature, enterprise management tooling, or a final availability timetable.

Three practical takeaways follow from Waymo’s published status:

Access is phased, beginning with a limited group of riders in named cities.
The product remains in beta, with the interface and experience still subject to enhancement.
Deployment claims need precision, because a gradual rider rollout is not equivalent to universal availability.

For mobility leaders, the key question is how a promising in-cabin AI beta can become a governed, useful customer service. Scalevise helps teams assess AI strategy, operating controls and implementation priorities before customer-facing capabilities scale. A clear plan can distinguish an early pilot from an operationally ready program while keeping decisions tied to verified product status. Discuss an AI consultancy with Scalevise to map the next practical step.

Frequently Asked Questions

What is the current status of Gemini in Waymo’s Ojai?

Waymo describes Gemini in the Ojai cabin as a beta feature with a redesigned user interface and ongoing enhancements.

Who received access in Waymo’s initial Ojai rollout?

Waymo said initial riders would be welcomed in San Francisco, Phoenix and Los Angeles, with free rides for a limited group while the company collected feedback.

Is Gemini in Ojai available to every Waymo rider?

No. The official May and July 2026 descriptions characterize the rollout as limited, gradual and still in beta rather than universally available.

What changed between Waymo’s May and July 2026 updates?

The May update outlined early rider access and feedback collection. The July update described Gemini in the Ojai cabin as a beta feature with a redesigned UI and ongoing enhancements.

What Gemini capabilities has Waymo confirmed for Ojai?

The supplied official descriptions confirm a Gemini beta in the Ojai cabin, but do not specify a complete list of supported interactions or controls.

Conclusion

Waymo’s Gemini integration in Ojai is a real but still limited deployment. The company’s updates point to a deliberate progression from a small rider group to gradual expansion, with product changes continuing during beta. The most accurate reading is not a universal launch, but an in-cabin AI experience that Waymo is actively testing and refining.

OpenAI Rolls Out Flexible Codex Pricing for Business and Enterprise Teams

Ali Farhat — Fri, 21 Aug 2026 21:00:30 +0000

OpenAI has formally rolled out a more flexible way for organizations to buy and use Codex. Eligible ChatGPT Business and Enterprise workspaces can add Codex-only seats on a pay-as-you-go basis, with usage billed through API-style token consumption rather than a fixed fee for each of those seats.

The change matters because it separates AI coding access from a conventional per-user subscription model. A team can provision Codex-only members for development work while managing consumption through credits, rather than treating every user as a full ChatGPT seat. OpenAI also reduced the annual ChatGPT Business seat price from $25 to $20 and introduced promotional credits intended to lower the cost of onboarding new Codex users.

OpenAI details the model in its official flexible Codex pricing announcement for teams. The initial announcement was made on April 2, 2026. A June 24, 2026 update clarified that existing pay-as-you-go seats remain in place, while the availability of new Business pay-as-you-go seats may change.

How OpenAI's Codex team pricing works

The central change is the availability of Codex-only seats within Business and Enterprise workspaces. These seats have no fixed seat fee. Instead, their Codex consumption is charged using API-style token rates. OpenAI's approach gives organizations a way to make Codex available to users whose primary need is AI-assisted coding, without requiring the same seat model used for standard ChatGPT access.

This is not a claim that coding costs are universally lower. A pay-as-you-go model makes expenditure more directly dependent on usage, so the financial effect will vary by team and workload. The documented advantage is billing flexibility: organizations can align Codex costs with token consumption and use credits across supported workspace features where their plan allows it.

Item	ChatGPT Business seat	Codex-only seat
Pricing approach	$20 per seat annually, reduced from $25	Pay as you go through API-style token consumption
Fixed seat fee	Per-seat annual price	No fixed fee for Codex-only seats
Onboarding promotion	Not specified in the announcement	$100 in Codex credits per new member, capped at $500 per team
Workspace context	ChatGPT Business	Eligible ChatGPT Business and Enterprise workspaces

Shared credits change the purchasing model

OpenAI's pricing framework also links usage across supported workspace tools. Credits and usage for Codex, ChatGPT Work, ChatGPT for Excel, and related Workspace features can draw from a common pool when that capability is available on the plan. OpenAI Help Center guidance corroborates that credits can work across these supported features and that Codex usage follows API-style token pricing.

For administrators, that creates a broader budgeting question than the cost of a single coding tool. The same credit pool can cover several AI-enabled workflows, which may simplify purchasing but also means usage in one feature can affect the credits available to another. Organizations will need to understand which features their plan supports and how they want to allocate credits across development and business use cases.

The rollout includes several practical elements:

Codex-only members can be added without fixed seat fees, with consumption billed by token usage.
New Codex-only team members are eligible for $100 in promotional Codex credits, up to $500 per team.
ChatGPT Business annual seats have a documented price reduction from $25 to $20 per seat.
Supported workspace tools can use a shared credits model, rather than keeping credits isolated to Codex alone.
Existing pay-as-you-go seats remain unaffected by the June update, although new-seat availability for Business may change.

What developers and enterprise teams should watch

For developers, Codex-only seats could make it easier for a company to extend AI coding access to targeted groups. The operational trade-off is that usage-based billing requires closer attention to consumption than a simple fixed-seat model. Teams should account for token use as part of their engineering-tool budget rather than assuming a predictable monthly cost per user.

For Business and Enterprise administrators, the shared-credit design is equally important. Codex spending cannot be viewed entirely in isolation if credits are also used for ChatGPT Work, Excel, or related features. The relevant question becomes how a workspace governs access and usage across its supported AI tools.

OpenAI's June 24 clarification also makes availability worth monitoring. The company states that existing pay-as-you-go seats are unaffected, but it notes that new pay-as-you-go seats for Business may be subject to changing availability. Organizations considering the model should therefore confirm the options visible in their own workspace before planning a broader rollout.

For organizations evaluating AI coding access, the practical issue is not simply token cost. It is how shared credits, workspace permissions, and developer workflows fit into governance and budgeting. Scalevise helps teams map AI tooling to measurable operating needs and implementation controls through its AI consultancy services. Request a consultation to assess a Codex and ChatGPT workspace approach before expanding usage.

Frequently Asked Questions

What is OpenAI's pay-as-you-go Codex pricing for teams?

It is a pricing model for Codex-only seats in eligible ChatGPT Business and Enterprise workspaces. Those seats have no fixed fee, and Codex usage is billed through API-style token consumption.

Which workspaces can use Codex-only pay-as-you-go seats?

OpenAI's rollout targets eligible ChatGPT Business and Enterprise workspaces. The company has also said that availability of new pay-as-you-go seats for Business may change, while existing seats remain unaffected.

How does the Codex promotional credit offer work?

OpenAI offers $100 in Codex credits for each new Codex-only team member. The promotion is capped at $500 per team.

Are Codex credits separate from ChatGPT Work and ChatGPT for Excel credits?

Where supported by an eligible plan, credits and usage for Codex, ChatGPT Work, ChatGPT for Excel, and related Workspace features can draw from a common pool.

Did OpenAI change ChatGPT Business pricing?

Yes. OpenAI documented a reduction in the annual ChatGPT Business seat price from $25 to $20 per seat.

Conclusion

OpenAI's flexible Codex pricing gives eligible Business and Enterprise workspaces a consumption-based option for AI coding access, while linking supported AI tools through shared credits. The lower ChatGPT Business seat price and limited onboarding credits add incentives, but the value of the model will depend on how each organization manages usage, access, and the availability of new seats.

OpenAI GPT-5.6 Pricing Update Cuts Terra and Luna Costs, Leaves Sol Unchanged

Ali Farhat — Fri, 21 Aug 2026 20:45:30 +0000

OpenAI has updated the price-performance positioning of its GPT-5.6 model family, cutting prices for the Terra and Luna tiers while keeping flagship Sol pricing unchanged. For API teams, the distinction matters: lower-cost model options can reduce routine inference spend, but organizations using Sol should not budget for a broad temporary reduction in its standard token rates.

According to OpenAI's GPT-5.6 price-performance update, Luna pricing was reduced by 80% and Terra pricing by 20%. The update also adds a Fast mode for Sol that can provide up to 2.5 times faster speeds at twice the price. It is a throughput option, not a standard-price discount for Sol.

At launch, OpenAI listed GPT-5.6 Sol at $5 per million input tokens and $30 per million output tokens. Terra was listed at $2.50 for input and $15 for output, while Luna was listed at $1 for input and $6 for output. The later update changes the economic case for Terra and Luna, but does not change Sol's stated standard pricing.

What changed in the GPT-5.6 model lineup

The update reinforces a tiered approach to model selection. Sol remains the flagship tier. Terra is the lower-cost option, and Luna is positioned as the fastest and most affordable tier. Rather than lowering every model's price, OpenAI has made the lower-priced tiers more economical and introduced an explicit premium path for faster Sol execution.

GPT-5.6 tier	Launch pricing per million input tokens	Launch pricing per million output tokens	Update described by OpenAI
Sol	$5	$30	Standard pricing unchanged; Fast mode offers up to 2.5x faster speeds at twice the price
Terra	$2.50	$15	Pricing reduced by 20%
Luna	$1	$6	Pricing reduced by 80%

For developers, this creates a clearer decision between capability, cost, and latency. Teams should evaluate model assignment at the workload level rather than assume a family-wide price change. In particular, the update supports revisiting which requests truly require Sol and which can be handled by Terra or Luna.

Relevant operational questions include:

Cost sensitivity: High-volume tasks may benefit most from the reduced Terra and Luna pricing.
Latency requirements: Sol Fast mode may suit time-sensitive workloads, but its stated price is twice the standard Sol rate.
Model routing: Applications that can direct requests by task type may be better positioned to use the lower-cost tiers.
Budget controls: Finance and platform teams should treat Sol Fast mode as a separate premium consumption option in forecasts and monitoring.

Budgeting and governance implications for API teams

The pricing changes make model governance more important, not less. Lower token costs can improve the economics of experimentation and scale, but they can also obscure where spending is accumulating if teams switch models or adopt faster modes without clear policies.

A practical governance approach starts with maintaining a model inventory: which applications use Sol, Terra, or Luna, and for what type of work. Teams can then set workload-specific expectations for quality, latency, and spending. This is especially useful when a platform offers both lower-cost tiers and premium speed options inside the same model family.

The update also illustrates why organizations should separate published standard pricing from temporary promotions, performance modes, and model-specific changes. Sol's standard input and output prices remain unchanged in OpenAI's update. Its new Fast mode changes the cost-speed trade-off, whereas the reductions apply to Terra and Luna. Treating those as interchangeable could lead to inaccurate cost projections.

For businesses building AI-powered customer experiences or internal tools, affordability is only one part of procurement and platform decisions. Model selection still requires testing against the task at hand, along with controls for usage visibility and approval of higher-cost execution paths.

As OpenAI pricing and model options evolve, businesses need a clear view of how their brands appear in AI-generated answers and where model-driven discovery affects demand. Scalevise helps teams measure that exposure with an AI Visibility and GEO Checker, turning AI search presence into actionable insight for content and growth planning. Start an AI Visibility scan to identify the questions, competitors, and answer surfaces that deserve attention.

Frequently Asked Questions

Did OpenAI reduce GPT-5.6 Sol API pricing?

No. OpenAI's GPT-5.6 update states that Sol pricing remains unchanged. The price reductions apply to Terra and Luna.

What are the listed GPT-5.6 Sol token prices?

At launch, OpenAI listed Sol at $5 per million input tokens and $30 per million output tokens.

Which GPT-5.6 tiers received price reductions?

OpenAI states that Luna pricing was reduced by 80% and Terra pricing by 20%.

What is GPT-5.6 Sol Fast mode?

Sol Fast mode is an option that OpenAI says can deliver up to 2.5 times faster speeds at twice the price. It is not a reduction in standard Sol pricing.

Conclusion

OpenAI's GPT-5.6 pricing update improves the cost profile of Terra and Luna while preserving Sol's standard rates. The addition of Sol Fast mode gives teams another performance option, but at a stated premium. API buyers should update model-routing, budgeting, and governance practices around the specific tier and execution mode they use.

Google Gemini Notebook Expands Into AI Mode Search With Cross-App Notebook Syncing

Ali Farhat — Fri, 21 Aug 2026 20:30:30 +0000

Google has renamed NotebookLM to Gemini Notebook, positioning its source-grounded research tool more tightly within the Gemini ecosystem. The change is accompanied by notebook syncing across Gemini Apps, Gemini Notebook, and AI Mode in Google Search, giving users more ways to create, organize, and revisit research projects without treating each interface as a separate workspace.

The rebrand is more than a logo change. According to Google's Gemini Notebook announcement, the service remains a standalone research tool but is intended to do more across Google's products. Google is also rolling out an upgrade that assigns each notebook a secure cloud computer for native code execution. That capability is initially available to Google AI Ultra users and eligible Google Workspace customers, with a planned expansion to Pro users in the following weeks.

For enterprises already using Google Workspace and Gemini, the important development is the connected notebook workflow. Teams can keep projects, conversations, and source collections associated with a notebook while moving between Gemini interfaces. But the rollout is broad rather than universally identical at every moment: Google describes phased access for cloud-enabled capabilities, and mobile access may require users to update the relevant app.

A research workspace that now spans Google interfaces

Google's documentation describes notebooks as shared objects across Gemini Apps, Gemini Notebook, and the Gemini app. Users can create, edit, and sync notebooks between those interfaces, with their associated sources remaining accessible. In practical terms, this lets a research effort begin in one Gemini surface and continue in another without recreating the source set.

AI Mode in Google Search adds a search-led entry point. Users can create and manage notebooks within AI Mode, and those notebooks automatically sync across AI Mode, Gemini, and Gemini Notebook. This matters because search research, document-based analysis, and conversational assistance often occur in separate tools. Google's approach connects those activities around the notebook rather than around a single chat thread.

The AI Mode sync boundary matters

The integration is not a complete two-way record of every interaction. Google's help documentation says notebooks sync across the three surfaces, but Gemini and Gemini Notebook chats do not sync back to AI Mode. That distinction is significant for teams that need to understand where a conclusion originated and which interactions form part of the accessible record.

Workflow element	AI Mode in Google Search	Gemini and Gemini Notebook
Create and manage notebooks	Supported	Supported across Gemini interfaces
Notebook syncing	Syncs with AI Mode, Gemini, and Gemini Notebook	Shares notebooks and sources across the connected workspace
Chat syncing back to AI Mode	Does not receive Gemini or Gemini Notebook chats	Gemini and Gemini Notebook chats do not sync back to AI Mode

This is not necessarily a limitation for every use case, but it should shape process design. A team using AI Mode for discovery and Gemini Notebook for deeper source work should not assume that the AI Mode view provides a full audit trail of later Gemini conversations. Where provenance, review, or retention requirements apply, teams should determine which interface is the system of record for a given workflow.

Secure cloud computers extend the notebook's role

Google's new cloud-enabled capability gives each notebook a secure cloud computer that can run code natively. This extends Gemini Notebook beyond organizing and reasoning over sources into tasks that may require computation. Google has described the initial availability as limited to AI Ultra users and Google Workspace customers with AI Ultra Access or AI Expanded Access, followed by a broader rollout to Pro users.

That staged availability means organizations should separate two decisions: adopting the connected notebook workflow and enabling cloud-based code execution. The first is tied to the wider rename and integration rollout. The second depends on a customer's plan and Google's phased release schedule. Treating them as distinct capabilities can help technology leaders avoid assuming that all users receive the same functionality simultaneously.

Governance questions for enterprise teams

Google emphasizes source-grounded responses and privacy terms for notebook content. Those attributes are relevant to enterprise governance, but they do not remove the need for internal controls. The new cross-interface model increases the number of places where users can initiate work on the same notebook, making clear ownership and operating practices more important.

Organizations evaluating Gemini Notebook should focus on three practical questions:

Source stewardship: Which users can add, change, or rely on sources within a shared research notebook?
Workflow traceability: Which interface should employees use when a research process requires a reviewable record, given the chat-sync boundary with AI Mode?
Feature eligibility: Which users have access to the cloud computer capability, and does that access align with internal requirements for code execution?

The broader strategic implication is that Google is making notebooks a common layer between search and Gemini. Rather than competing as an isolated note-taking destination, Gemini Notebook is becoming a workspace that carries context across Google AI products. Its value for a business will depend less on the new name than on whether teams can establish repeatable, governed ways to collect sources, conduct analysis, and share outcomes.

For businesses, connected AI research can reduce context switching, but it also makes fragmented ownership and unclear data-handling practices more costly. Scalevise can help assess where Gemini Notebook and AI Mode fit within your existing research, governance, and AI adoption processes, then define practical controls for the workflows that matter most. Explore our AI consultancy for governed AI implementation and request a consultation.

Frequently Asked Questions

What is Gemini Notebook?

Gemini Notebook is Google's new name for NotebookLM. Google says it remains a standalone research tool while becoming more integrated across Google products.

Can Gemini Notebook notebooks be used in AI Mode in Google Search?

Yes. Google says users can create and manage notebooks in AI Mode, with notebooks syncing across AI Mode, Gemini, and Gemini Notebook.

Do Gemini chats sync back into AI Mode?

No. Google states that chats from Gemini and Gemini Notebook do not sync back to AI Mode, even though the notebooks themselves sync across the connected interfaces.

Who can use Gemini Notebook's cloud computer capability?

Google says the native code execution capability is initially available to Google AI Ultra users and eligible Google Workspace customers with AI Ultra Access or AI Expanded Access. A rollout to Pro users is planned in the following weeks.

Conclusion

Google's Gemini Notebook rollout makes the former NotebookLM a more connected part of its AI ecosystem, linking source-based research with Gemini and AI Mode in Google Search. The shared notebook model can make research workflows more continuous, while the chat-sync boundary and phased cloud-computer access give enterprises concrete details to consider before standardizing how teams use the tool.

GA4 Misattributed 22.4% of AI Overview Events as Direct in a Nine-Month Study

Ali Farhat — Fri, 21 Aug 2026 17:45:30 +0000

A nine-month, first-party GA4 study has identified a material reporting problem for teams measuring Google AI Overviews. Of 51,200 tracked AI Overview events for one brand, 11,468 events, or 22.4%, were attributed to Direct rather than Organic Search. The result suggests that conventional acquisition reporting can understate the contribution of AI Overview visibility to organic traffic.

The figures come from Search Engine Land's reporting on the 51,000-event AI Overview study, which describes data collected from September 2025 to June 2026. The analysis is not a market-wide benchmark. It covers one brand and depends on a custom approach for identifying AI Overview visits. Still, it offers a documented example of how AI-driven search journeys can create gaps between user acquisition and the channel ultimately credited in analytics.

For enterprise teams, the issue is larger than a dashboard classification error. Channel attribution informs SEO investment, content priorities, performance targets and executive reporting. If a meaningful share of AI Overview-driven visits arrives in GA4 as Direct, a business could conclude that organic search is delivering less value than it actually is.

Why AI Overview attribution can distort organic reporting

The study estimated that AI Overview traffic accounted for about 7.5% of the brand's organic sessions across the full measurement window. Reported rounded estimates differed slightly, at 7.53% and 7.46%, but both point to the same overall finding: AI Overview activity represented a measurable portion of organic traffic for this site. That share also varied substantially month to month.

The attribution issue varied too. The reported monthly Direct misattribution rate ranged from roughly 16.8% to 29.3%, rather than remaining fixed at 22.4%. That variability matters because a single-period report may either understate or overstate the problem for a given month. Teams should resist treating one percentage as a universal correction factor for every website, market or reporting period.

The dataset also recorded 1,661 cited AI Overview snippets. The most frequently cited snippet was associated with around 2,276 events. Citation activity can help explain where a brand appears in AI Overview results, but it is not equivalent to a clean measurement of referral or organic session attribution. Visibility, clicks, sessions and conversion behavior are related signals, not interchangeable metrics.

Measurement approach	What it can show	Important limitation
Standard GA4 channel reporting	How visits are classified in existing acquisition funnels	AI Overview-driven visits may be credited to Direct instead of Organic Search
Custom AI Overview identification using URL fragments	Events associated with the study's AI Overview detection method	The text-fragment identifier is not globally unique, which can affect exact counts
Cross-referencing analytics and visibility signals	A fuller view of AI-driven visibility, landing-page behavior and reported search activity	It requires validation across multiple datasets rather than reliance on one channel label

The measurement caveat is central, not incidental

The researchers used a custom dimension and fragment-based method to identify AI Overview traffic. That methodology is valuable because it surfaces activity that a standard GA4 view may not make obvious. However, the text-fragment identifier in AI Overview URLs is not globally unique. As a result, exact event counts may be affected, and other organizations should not assume they can reproduce the figures simply by applying the same rule.

GA4 configuration may also influence outcomes. The study therefore supports a specific conclusion: one documented implementation found a non-trivial Direct attribution gap for AI Overview events. It does not establish that every GA4 property misattributes 22.4% of this traffic, nor does it prove a single cause for every Direct-classified visit.

What enterprise analytics teams should do now

The immediate priority is not to rewrite historical channel reports with an assumed adjustment. It is to test whether a similar gap exists in the organization's own data and document the method used. A defensible process should bring together acquisition data, search visibility evidence and on-site behavior rather than treating any one signal as conclusive.

Useful steps include:

Audit Direct traffic patterns on landing pages that are frequently associated with AI Overview citations or organic search demand.
Create a documented detection method for relevant AI Overview visits, while recording its assumptions and known limitations.
Compare GA4 findings with Search Console signals, citation tracking and landing-page engagement to identify inconsistencies worth investigating.
Separate observed data from inferred attribution in executive reports, particularly when measuring SEO performance or AI search visibility.
Review reporting governance regularly because month-to-month variation means the issue may not be stable over time.

This approach is especially important where channel data informs budget allocation or performance incentives. A reporting process that treats Direct as a homogeneous bucket can hide multiple user journeys, including visits that began with search but were not credited to Organic in the final analytics record.

For businesses building an AI search measurement program, the practical goal is a more reliable visibility baseline, not false precision. Scalevise can help teams connect AI search presence with the reporting signals that shape content and growth decisions through its AI Visibility and GEO Checker. A structured assessment can reveal where AI-generated search results, analytics attribution and organic reporting may diverge, so leaders can prioritize validation before changing strategy. Start an AI Visibility scan.

Frequently Asked Questions

What did the AI Overview GA4 study find?

The study tracked 51,200 AI Overview events for one brand from September 2025 to June 2026 and found that 11,468 events, or 22.4%, were attributed to Direct rather than Organic Search in GA4.

Does the 22.4% figure apply to every website?

No. The analysis covered a single brand and used a custom dimension and fragment-based method. Its result is a documented example of an attribution gap, not a universal GA4 benchmark.

How much organic traffic did AI Overviews represent in the study?

AI Overview traffic was estimated at about 7.5% of the brand's organic sessions across the full measurement period, with substantial month-to-month variation.

Why should SEO and analytics teams investigate Direct traffic?

If AI Overview-driven visits are classified as Direct, standard acquisition reports can undercount organic search's contribution and affect SEO reporting, investment decisions and performance analysis.

What is a sensible way to validate AI Overview traffic?

Cross-reference GA4 data with Search Console, landing-page behavior and citation tracking, while documenting the limitations of any custom AI Overview detection method.

Conclusion

The nine-month study shows that AI Overview measurement can introduce a meaningful attribution gap in GA4. Its single-brand scope and methodology limitations require caution, but the 22.4% Direct classification finding is strong reason for organizations to validate how AI-driven search visits appear in their own reporting. Better governance starts with treating channel labels as evidence to investigate, not as unquestioned proof of user origin.

Microsoft MAI-Image-2.6 Reaches No. 2 on Arena as Foundry Preview Begins

Ali Farhat — Fri, 21 Aug 2026 17:30:30 +0000

Microsoft has launched MAI-Image-2.6, the latest version of its first-party image-generation model. The release moves the model to No. 2 on Arena's text-to-image leaderboard and No. 3 on its image-editing leaderboard, according to Microsoft's MAI-Image-2.6 launch announcement. It is available in MAI Playground and in private preview through Microsoft Foundry, extending Microsoft's effort to bring proprietary AI models into its wider product ecosystem.

The key change is not simply a new model label. Microsoft reports a 79-point overall Elo gain over MAI-Image-2.5 on Arena, alongside a 91-point improvement in text rendering. The company also cites quality gains for portraits, 3D imagery, and polished commercial or photorealistic outputs. For teams producing product visuals, branding assets, or cinematic concepts, those categories are often where generated images must meet a higher bar for usability.

What Microsoft says changed in MAI-Image-2.6

MAI-Image-2.6 follows MAI-Image-2 and MAI-Image-2.5 in Microsoft's continuing image-model release cycle. Microsoft positions the new version as an improvement in generation and editing quality, with particular emphasis on control and output refinement rather than a wholesale change to the company's image-AI strategy.

The company highlights several areas of progress:

Stronger text rendering, a persistent challenge for image models used in marketing, product, and presentation work.
Higher-quality portraits and 3D imagery, alongside improvements to photorealistic and commercial-looking output.
Better grounding and control, including continuing work on multi-reference grounding, formatting and resolution, and cross-model consistency.
Improved Arena results, with Microsoft reporting No. 2 in text-to-image and No. 3 in image editing.

Arena rankings are a useful external signal of comparative model preference, but they should not be treated as a complete enterprise evaluation. A leaderboard result does not by itself answer how a model will perform against a company's brand rules, reference materials, approval processes, or deployment requirements. Those questions still depend on the specific workflow and the controls available in the platform where the model is used.

Area	MAI-Image-2.5	MAI-Image-2.6
Arena overall performance	Baseline for Microsoft's comparison	Microsoft reports a +79 Elo gain
Text rendering	Baseline for Microsoft's comparison	Microsoft reports a +91 improvement
Highlighted quality areas	Not detailed in the launch comparison	Text, portraits, 3D imagery, commercial and photorealistic outputs
Availability stated in the launch announcement	Not specified	MAI Playground and private preview on Microsoft Foundry

Availability, pricing, and platform rollout

MAI-Image-2.6 is currently usable through MAI Playground and in private preview on Microsoft Foundry. Microsoft says a broader rollout across Foundry and other products is planned in the near term. Its stated longer-term distribution strategy includes first-party AI capabilities across products such as Foundry, Copilot, and Bing Image Creator.

The announcement does not provide public pricing for MAI-Image-2.6 or a general-availability date for Foundry. That matters for enterprise planning: a private preview can help organizations assess a model and integration path, but it is not the same as a broadly available production service with published commercial terms. Teams should distinguish current access from planned platform availability when setting timelines or procurement expectations.

Why the release matters for enterprise image generation

The model's reported text-rendering improvement is especially relevant because many business image workflows need more than aesthetically appealing visuals. Product mockups, campaign assets, signage concepts, and branded graphics can fail review when embedded text is inaccurate or unreadable. Better results could reduce rework, although Microsoft has not published workflow-specific benchmarks or guarantees in the launch announcement.

Microsoft's emphasis on grounding, multiple references, formatting, resolution, and consistency also points to the practical requirements of business creative work. Organizations typically need predictable use of approved references and output that can move through existing review processes. The announcement signals ongoing work in those areas, but it does not detail the precise controls, governance features, or policy settings that will accompany every rollout surface.

For Microsoft customers, the strategic significance is the model's position within a larger stack. Rather than treating image generation as a standalone creative tool, Microsoft is building a progression of MAI models that can be distributed through its AI platforms and products. That could give enterprises a more direct route to evaluating a first-party model within the Microsoft environment they already use, once access broadens.

For businesses adopting generative imagery, model quality is only one part of the decision. They also need a clear approach to reference handling, human review, brand standards, and the platform's available controls. Scalevise's AI consultancy helps organizations turn emerging model capabilities into governed workflows that fit real business processes, helping teams evaluate where image AI can create value before deployment complexity grows. Request a consultation.

Frequently Asked Questions

What is Microsoft MAI-Image-2.6?

MAI-Image-2.6 is Microsoft's latest first-party AI image model. Microsoft says it improves on MAI-Image-2.5 in Arena performance, text rendering, and several image-quality categories.

Where is MAI-Image-2.6 available?

Microsoft says MAI-Image-2.6 is available in MAI Playground and in private preview on Microsoft Foundry. A wider rollout across Foundry and other products is planned in the near term.

How did MAI-Image-2.6 perform on Arena?

Microsoft reports that MAI-Image-2.6 ranks No. 2 on Arena's text-to-image leaderboard and No. 3 on its image-editing leaderboard. The company reports a 79-point overall Elo gain over MAI-Image-2.5.

Has Microsoft announced pricing for MAI-Image-2.6?

No public pricing is stated in Microsoft's launch announcement. The release identifies MAI Playground access and a private preview on Microsoft Foundry, but does not provide commercial terms.

Conclusion

MAI-Image-2.6 gives Microsoft a stronger reported image-model result and a clearer path into its Foundry ecosystem. Its improvements in text rendering, visual quality, and control-related areas are relevant to enterprise creative workflows, while private-preview status and undisclosed pricing mean organizations should treat the release as an evaluation opportunity rather than assume broad production availability today.

Microsoft Expands MAI Playground With Image, Voice, Transcription and Reasoning Models

Ali Farhat — Fri, 21 Aug 2026 17:15:31 +0000

Microsoft has expanded MAI Playground as a limited-preview environment for trying models from its MAI family, including image generation, speech, transcription and reasoning capabilities. The development gives users a public place to test Microsoft-built models while the company takes a staged approach to API and broader developer availability.

The clearest immediate use case is image generation. In its official MAI-Image-2 announcement, Microsoft said the model is being released and can be tried in MAI Playground. The company also said select customers can access the model through an API, with Foundry availability planned for developers later. WPP is identified as an early adopter.

That makes Playground more than a demonstration page. It is a public testing layer in Microsoft’s rollout path for MAI models, although it is not presented as a universally available production API. The Playground itself describes the service as a limited preview and warns that AI can make mistakes. Businesses evaluating the models should treat outputs as experimental results, not as evidence that a capability is ready for every production workflow.

What MAI Playground currently offers

Microsoft’s Playground lists a broader set of MAI models than the original MAI-Image-2 announcement alone. The current lineup spans several AI task types, allowing users to assess different parts of Microsoft’s emerging model portfolio in one environment.

Model	Capability indicated by its name and Playground listing	Availability context in the supplied sources
MAI-Image-2.5	Image generation	Listed in MAI Playground
MAI-Image-2.6	Image generation	Listed in MAI Playground
MAI-Transcribe-1.5	Transcription	Listed in MAI Playground
MAI-Voice-2	Voice	Listed in MAI Playground
MAI-Thinking-1	Reasoning	Listed in MAI Playground

The naming differences matter. Microsoft’s announcement specifically discusses MAI-Image-2, while the Playground lists MAI-Image-2.5 and MAI-Image-2.6. The supplied sources establish that the image family is being showcased through Playground, but they do not provide a feature-by-feature specification or benchmark comparison among those image variants. Users should therefore avoid inferring performance differences from version numbers alone.

The practical value of a multi-model playground is early evaluation. Teams can test whether a model category fits a workflow before making architecture decisions around API integration or a broader platform rollout. That is particularly relevant where a company needs to assess outputs, prompt behavior and operational controls across different media types.

Microsoft’s current access approach can be summarized as follows:

MAI Playground is available as a limited preview for hands-on testing.
MAI-Image-2 API access is available to select customers, rather than broadly described as open access.
Foundry integration is planned to open to developers later.
WPP is named as an early adopter for the image model.
The supplied sources do not state a public price for Playground use, MAI-Image-2 API access or future Foundry access.

This staged release distinguishes MAI Playground from a fully documented, generally available developer platform. It can help organizations explore the technology, but procurement, production deployment and cost planning require details that have not yet been published in the supplied material.

Enterprise governance and rollout implications

For enterprise AI teams, the important story is not only that Microsoft is exposing more MAI models. It is also that the company is placing the early experience inside a preview environment with stated limitations and policy controls.

Voice features illustrate why governance cannot be treated as an afterthought. Related MAI-Voice materials state that custom voice creation requires approval under Microsoft’s Responsible AI policies. That requirement is a meaningful signal for organizations considering voice-related use cases, where identity, consent, impersonation and brand risk can be material concerns.

The same discipline applies across the Playground’s model categories. Image generation can introduce review requirements around creative assets and intended use. Transcription workflows can involve sensitive audio or business records. Reasoning models may be useful for exploration, but the Playground’s warning that AI can make mistakes reinforces the need for human review and clear limits on automated decisions.

MAI Playground also reflects Microsoft’s gradual move from model experimentation toward developer distribution. The official announcement separates current trials, selected API customers and later Foundry availability. That sequence suggests that enterprises should distinguish three questions that are often incorrectly combined: whether a model can be tested, whether it can be integrated, and whether it can be governed at production scale.

For businesses planning evaluations, a useful approach is to define the intended task, the review process and the data boundaries before trial activity begins. Preview access may be valuable for comparing fit, but it does not remove the need for approval paths, ownership and testing criteria.

For businesses assessing new AI platforms, early access is only useful when it connects to a clear operating model for risk, integration and measurable outcomes. Scalevise helps teams turn exploratory model testing into a practical adoption plan through AI consultancy and implementation guidance, covering use-case selection, governance and workflow design. This is particularly important when preview tools evolve faster than internal controls. Request a consultation to discuss your AI implementation priorities.

Frequently Asked Questions

What is Microsoft MAI Playground?

MAI Playground is Microsoft’s limited-preview environment for trying models from its MAI family. The Playground says that AI can make mistakes, so it should be treated as a testing environment rather than an assurance of production-ready results.

Which models are listed in MAI Playground?

The supplied research identifies MAI-Image-2.5, MAI-Image-2.6, MAI-Transcribe-1.5, MAI-Voice-2 and MAI-Thinking-1 as the current Playground lineup.

Is MAI-Image-2 available through an API?

Microsoft says MAI-Image-2 API access is available for select customers. The company also says Foundry availability will open to developers later.

How does Microsoft govern custom MAI voice creation?

Related MAI-Voice materials state that custom voice creation requires approval under Microsoft’s Responsible AI policies.

Has Microsoft announced MAI Playground pricing?

The supplied sources do not state public pricing for MAI Playground, MAI-Image-2 API access or planned Foundry availability.

Conclusion

Microsoft’s MAI Playground gives users a limited-preview venue to evaluate a growing set of in-house models, led by the MAI image family and supported by voice, transcription and reasoning offerings. Its staged access model and Responsible AI controls are as important as the model list: organizations can begin testing now, while broader developer availability and public pricing details remain to be clarified.

SEO in 2027: Why AI Answer Visibility Will Matter Beyond Traditional Rankings

Ali Farhat — Fri, 21 Aug 2026 14:00:30 +0000

SEO is moving beyond the question of where a page ranks. By 2027, the more useful measure of search performance may be whether a brand is present and accurately represented across AI-generated answers, AI Overviews, zero-click results, and other discovery surfaces. That does not make traditional rankings irrelevant, but it changes what enterprises must measure, govern, and improve.

The shift is grounded in an evolving search environment rather than a single platform change. In Search Engine Land's analysis of the emerging brand-signal authority model, the publication outlines a 2027-facing view of SEO in which authority increasingly depends on entity-centric signals, brand mentions, and credibility across multiple sources. Its proposed direction is clear: links and rankings alone are less complete proxies for whether an AI system will surface a business.

For enterprise teams, the implication is practical. A top organic result can still fail to produce meaningful visibility if users receive an AI answer without encountering the brand, if the brand is omitted from an answer's cited or synthesized sources, or if inconsistent information weakens the signals systems use to understand the company.

From position tracking to AI-visible authority

Traditional SEO governance has typically centered on rankings, organic traffic, backlinks, crawl health, and page-level performance. Those controls remain valuable because search engines and AI systems still require discoverable, renderable, and indexable information. But they do not fully capture how AI-driven discovery assembles recommendations or summaries.

Search Engine Land's March 2026 analysis of the AI engine pipeline describes a 10-gate process spanning discovery, rendering, indexing, and later stages that determine whether a source can win visibility. The framework emphasizes why governance matters: signals can be lost, distorted, or weakened at several points before an AI-generated response is produced.

The April analysis extends that logic to authority. It predicts a move toward Share of Model, or SoM, as a visibility metric for understanding how often a brand appears in AI-model outputs. Whether or not SoM becomes a standard industry metric, the underlying issue is already relevant: organizations need a way to assess presence in answer systems, not simply page positions in conventional result pages.

SEO focus	Ranking-centric approach	AI-driven discovery approach predicted for 2027
Primary performance signal	Organic ranking position	Visibility within AI answers and broader discovery surfaces
Authority emphasis	Links as a major authority signal	Brand signals, entity understanding, mentions, and multi-source credibility
Measurement direction	Rankings and traffic reporting	Potential Share of Model measurement alongside established SEO metrics
Governance concern	Page and technical search performance	How signals persist through discovery, rendering, indexing, and AI recommendation processes

The comparison is not a reason to abandon ranking reports. It is a reason to treat them as one layer of a broader visibility model. Rankings describe a page's placement in a search result. They do not necessarily show whether an AI answer recognizes the underlying entity, trusts its information, or includes it when responding to a user.

What enterprise SEO teams should govern now

The strongest response is not to chase a presumed AI ranking factor. It is to create a more durable operating model for the signals an organization publishes and earns. Search Engine Land's analysis points to a future where multi-source authority matters more, making inconsistent brand information and disconnected ownership more costly.

A useful enterprise program should connect technical SEO, content, communications, legal, product, and data stakeholders. That is especially important where organizations manage large sites, multiple markets, regulated claims, or a broad ecosystem of third-party references.

Teams should prioritize several areas:

Entity clarity: Keep company, product, service, and subject information consistent enough for systems to identify relationships accurately.
Multi-source credibility: Treat reputable mentions and corroborating sources as part of authority, rather than viewing links as the only external signal that matters.
Content quality and maintenance: Publish clear, specific information and maintain it as products, policies, and positioning change.
Technical accessibility: Continue governing discovery, rendering, indexing, and crawl-related controls so important information can be processed as intended.
AI visibility measurement: Add recurring checks for how brands and key topics appear in AI answers, alongside established ranking and traffic reporting.

Crawl and data policies deserve particular attention, but not as a one-time compliance task. Organizations need clear ownership over the technical rules that affect access to their content, as well as an informed process for reviewing the business trade-offs of those rules. The available research does not establish a universal policy that every enterprise should adopt. It does establish that technical governance is part of a larger chain that influences discovery visibility.

This change also affects content strategy. A content calendar built solely around queries and page rankings can miss the wider question of whether the organization provides coherent, trustworthy information about the entities and topics it wants to own. The more AI systems synthesize material from multiple sources, the more important it becomes for a brand's first-party content, public references, and factual claims to align.

For business teams, the risk is not simply lower traffic. It is reduced visibility at the point where customers form an answer, shortlist providers, or validate a decision. Scalevise helps organizations turn this broader search challenge into measurable governance by identifying how their brand appears across AI discovery systems and where critical gaps exist. Use the AI Visibility GEO Checker to establish a clearer baseline, prioritize high-value issues, and start an AI visibility scan.

Frequently Asked Questions

What does SEO in 2027 mean for traditional rankings?

Traditional rankings are still useful, but they are unlikely to be a complete measure of search visibility. The emerging model adds visibility in AI-generated answers, AI Overviews, and other discovery surfaces.

What is Share of Model in SEO?

Share of Model, or SoM, is a 2027-facing metric proposed in Search Engine Land's analysis. It describes measuring how often a brand appears in AI-model outputs rather than relying only on conventional ranking positions.

Why are brand mentions and entity signals becoming more important?

The cited analysis predicts that AI-driven discovery will place greater weight on entity-centric understanding, brand mentions, and credibility across multiple sources. These signals can help systems understand and assess a brand beyond its backlink profile.

Should enterprises change their crawl and content policies for AI search?

Enterprises should review their technical access controls and content governance as part of a broader visibility strategy. The research supports stronger governance across discovery, rendering, and indexing, but it does not prescribe one universal policy for every organization.

Conclusion

The 2027 SEO outlook is not a declaration that rankings are obsolete. It is a shift toward a wider definition of visibility, one that includes whether AI systems can discover, understand, trust, and surface a brand. Enterprises that combine technical search discipline with entity clarity, multi-source credibility, and AI visibility measurement will be better positioned to manage that transition.